How do I clean up a pile of dashboards nobody uses?

Sprawl is a symptom, not the disease

Every duplicate dashboard exists because someone could not find or could not trust an existing one. If you delete the duplicates without fixing the trust problem, they grow back within a quarter under slightly different names.

So the audit has two outputs. One is a shorter list of reports. The other is a list of reasons people built their own, which is usually some combination of missing segments, wrong date basis, stale data and no drill path.

Ask the people who built their own what the canonical report was missing. The answers are consistent across companies and they are specific: it did not break out the segment I manage, it used a date basis that did not match my numbers, it was a day stale, or I could not get from the total to the jobs.

The audit table

One row per report, and the columns are what make the decision obvious.

• Owner. A person, not a department. No owner means no survivor.
• Decision supported. Written as a sentence describing an action someone takes because of it.
• Distinct viewers, last 90 days. Distinct matters; one person refreshing daily is not adoption.
• Last opened. Plus whether the only opens came from the person who built it.
• Sources and definitions used. This is how you find the reports that disagree with each other.
• Overlap. Which other report already answers most of this.

Consolidation rules that survive contact

One canonical report per audience level rather than per person: an ownership view, an operations view, a marketing view, a location view. Build for the role, so a new manager inherits reporting instead of commissioning it.

Exceptions and changes do not belong in new tabs. They belong in briefs and alerts, which is how most of the long tail disappears without anyone losing information they were actually using.

Give the long tail somewhere to go

Some one-off analyses are legitimate. They just should not live forever in the same space as governed reporting. A scratch area with an expiry date and no promise of maintenance handles them honestly, and it keeps the canonical set small enough to actually maintain.

When we rebuild reporting for a company, the count of reports almost always goes down. The measure of success is how many people open something in a week, not how many things exist to open.

Archive rather than delete on the first pass, and announce it. If nobody asks for a report within a month of it disappearing, the archive can be cleared. If two people ask within a day, you learned something the view logs did not tell you.